I’ve encountered the following slightly frustrating situation a few times in Live:
- I adjust/set picking parameters, and apply to all mics.
- The graph of NCC vs local power comes up, and I adjust thresholds and click “Extract all”
- I go back sometime later and realize that I am missing lots of particles! I then realize that because in step 2, Live had only seen a handful of mics, the total local power range (the min and max, which limit the range that I can set initially) is not representative of the entire dataset, and so on low defocus micrographs it is then missing lots of particles.
To avoid this situation, would it be possible to allow users to disable filtering by local power? Or set to just filter out the top/bottom x%?
Hey @olibclarke —
Just discussed this internally! Long story short: we’ve added this to our list of Live improvements for a future release.
An important note here is that we don’t set those thresholds based on absolute measure of power, but instead compute a ‘Power Score’ that is relative to a ‘calibration line’ which models variations of power as a function of defocus. If those early micrographs don’t span a sufficiently representative range of defocus values, then that linear fit will be off. We have an (old-ish) tutorial on Particle Picking Calibration that explains this in a bit more detail.
For the future, we like your idea of adding a percentile filtering that can operate before a sufficient calibration is possible.
Thanks for the feedback as always,